You retopologised it. Is it still the same shape?

You spend a day laying quads over a sculpt. It looks right in the viewport. Then you bake the normal map and the shoulder has a smear across it, the ear is soft, and one spot on the jaw is dark for no reason you can see. Now you have to guess: is the cage too tight, did a vertex get left behind, or did you simply never cover that fold at all?

RetopoProof answers that before you bake. It measures your new mesh against the original surface, in millimetres, and tells you which vertex. Then it measures the other way round — the original against your new mesh — because that is the only way to find something you did not cover at all.

Both directions, because one of them is blind

Every tool that checks a retopology measures from the new mesh to the old one. That direction cannot see a missing ear. Where a feature is absent, your retopology is not wrong: there is simply nothing of it there, and every vertex you do have sits perfectly on the surface. The number comes back clean and the ear is still gone.

So the original is measured back against the retopology as well. On a measured case with a sculpted lump the forward number was 0.00 mm — perfect — and the reverse measurement found two uncovered patches, 4.6 and 2.6 edges across, the deepest point 523 mm from anything in the new mesh.

The vertex that passes every tolerance and is still wrong

One vertex the snapping missed sat 0.393 mm off the surface, under a one-millimetre tolerance. RetopoProof named it anyway, because it does not compare that vertex to a threshold — it compares it to the rest of your own mesh, whose middle sits at 0.000 mm. The line falls at 0.192 mm, and past it this mesh has stopped being consistent with itself. That vertex is not a shape you chose. It is one that got left behind.

Where the line falls is measured, not decided. Comparing a vertex to its neighbours is meaningless when every distance is rounding noise: on twelve real production models with a snapped retopology, the worst vertex already sits seven to two hundred and forty times above the average, and a tool judging on that ratio alone shouts at ten of the twelve. So the line carries a floor taken from the mesh itself — a thousandth of its own edge length. Below that, nothing is claimed at all.

And it is measured from the middle of the mesh, not its average, for a reason that only shows up when the work is bad. Every vertex left behind drags the average up with it. Judged against the average, fifteen stragglers look milder than one, and twenty look like nothing at all — the tool goes silent exactly where it is needed most. Judged against the middle, fifteen out of three hundred move nothing, all fifteen are found, and the report says how many there are instead of saying "one".

Why it does not judge you in millimetres

A millimetre means nothing until you know how big the model is. A percentage of the model means nothing until you know how dense the mesh is: any polygon cuts the corner off a curve, and by how much depends on edge length against radius. A tool that picks a number and calls everything past it a defect will shout at every honest low-poly mesh on earth.

So the second half of the measurement is judged against your own mesh — its median edge length. A mesh cannot carry detail finer than its own face, and calling that a defect would be a lie. What it could carry, and does not, is reported.

That rule was not guessed, it was measured. Four densities of mesh over a noisy sculpt, a smooth sculpt, a hard-surface model, an honestly decimated organic form and eleven real production models: zero uncovered patches on every one of them. A tree with three million vertices: 368 specks of foliage, the largest 0.38 of an edge across — correctly reported as detail, not as loss. Real missing features: 2.0, 3.1, 4.6, 6.0 and 6.5 edges across. There is clean daylight between the two, and in millimetres there is none.

It is tested on being quiet, too

A diagnostic that always finds something is as useless as one that never does, so silence is tested as hard as noise — and each half of the measurement gets its own healthy sample, because they are not the same thing. A retopology snapped to the surface reads 0.00 mm, on all twelve real production models tried, with no stray vertex reported. A retopology of the same shape but coarser leaves nothing uncovered, on eleven of those twelve.

The twelfth is worth saying out loud. It is an assembly of thin pipes, and coarsening it that far genuinely destroys it: the new mesh runs beside the pipe rather than along it, 42.07 mm off at an edge length of 13.0 mm. RetopoProof reports it, and that is the right answer. A sample you demand silence on has to be a sample of finished work — otherwise you are not testing the tool, you are training it to lie.

It measures what your bake will see

• The evaluated mesh, not the file. A live Shrinkwrap, Subdivision or Mirror is part of the shape, so it is part of the measurement. If anything other than a shrinkwrap is shaping your retopology, that is said in its own line, because then the number describes the whole stack.
• World coordinates, all the way down. Object scale is part of the answer, and non-uniform scale is where naive tools break: finding the nearest point inside the object and only then converting to world gives a point on the surface that is not the nearest one in world at all. Measured on a mesh scaled (1, 1, 8) that mistake reports 199.62 mm where the truth is 54.05, and on (0.2, 1, 5) it reports 126.94 where the truth is 26.25 — two and a half to nearly four times too much, a pure generator of false alarms, on exactly the unapplied scale that scans and blockouts arrive with. Every distance here comes from a tree built on world-space triangles, and matches an independent world measurement to the third decimal at (1,1,1), (1,1,8), (0.2,1,5) and (100,100,100) alike.
• Inside versus outside. A vertex sunk inside the original bakes worse than the same distance outside, so those are counted separately — and only where the distance is real, because at zero distance the direction is just rounding noise.
• Big scans. A very dense mesh is sampled deterministically, and the number of points actually checked is printed, so you always know what was looked at. The sample is drawn by a mixed hash, not by taking every Nth vertex: vertex numbers in a scan run in rows, an every-Nth sample lands in neat columns, and a stripe missing between those columns is then never looked at. On a 400 by 400 scan that mistake reported full coverage over a gap 399 edges wide.
• It refuses rather than guesses. A mesh switched off in the viewport, or in a collection that is switched off or excluded, has no evaluated mesh at all — Blender simply does not build one. RetopoProof says so and stops, instead of failing halfway through. A mesh with vertices but no faces gets the same treatment: distance to a surface needs a surface.
• Numbers you can act on, or a warning that you cannot. When a live Subdivision or Mirror changes the vertex count, the worst vertex number belongs to the evaluated mesh and is not one you can select — so the report says that in the same sentence rather than sending you to a vertex that does not exist, for the original as well as the retopology.
• The mesh that gets baked, not the one on screen. A subdivision set to one level in the viewport and three in the render is the most common setup in Blender, and both of its switches are on — so nothing flags it, and every measurement you take describes a surface that will not be rendered. On a measured case that gap was 25.81 mm while the report read 0.00. RetopoProof compares the viewport and render settings of every modifier on both objects and says which ones disagree.
• Depth as well as width. A feature narrower than one face cannot be carried by the mesh — but a bolt recess 133 mm deep is not surface grain, and calling it that would be a lie. Anything deeper than two of your own edges is reported on its own, even when it is too narrow to be a patch.
• Patches, not speckles. A region only counts as uncovered when the points inside it are actually far — measured, as a share of everything looked at there. Even surface grain used to weld into one report-wide "missing patch"; now it is counted as grain, because half of what sits inside it is close by.

It changes nothing

RetopoProof reads. It does not snap, relax, project or fix. Every finding comes with one line of what to do and the number of the vertex to go to; whether a gap is a mistake or a decision is your call. A fingerprint of the session is taken before and after every measurement and compared, and the panel prints the result of that comparison every time — not only when something went wrong.

Tested

229 checks on each of Blender 4.2.9, 5.1 and 5.2, zero problems. Including sixteen deliberate sabotages of the product itself, to prove the tests can fail: all sixteen turn the suite red. The proof that nothing in your scene changed is itself tested by sabotage — moving a vertex, moving the object or clearing its materials during the measurement all make the fingerprint differ, so the guarantee is not one that cannot fail.

Requires

Blender 4.2 or newer. No internet connection, no external libraries, nothing written to your file system.